I understand that VarList is used to add variables on the fly in Pyomo. But I've never seen a VarList that allows adding variables "on-the-fly" with indices.
Below I've shown the actual and desired behavior:
import pyomo.environ as pe
### actual behavior
m = pe.ConcreteModel()
m.x = pe.VarList()
for i in range(2):
for j in range(2):
new_v = m.x.add()
print(list(m.x._index_set))
### desired behavior
m = pe.ConcreteModel()
m.x = pe.VarList()
for i in range(2):
for j in range(2):
new_v = m.x.add((i,j))
print(list(m.x._index_set))
### ideally this would print: [(0,0), (1,0), (0,1), (1,1)]
Such a feature would allow me to access the VarList in the future like so: m.x[i,j].
I've seen this question: Pyomo and VarList - generate variables automatically, but it doesn't use VarList in the answer, and therefore won't support adding variables on the fly.
The short answer is that you cannot create indexed VarList components. The reasons boil down to how Pyomo implements indexing sets: for an "indexed VarList," you really want a Set of independent Sets (which Pyomo supports) - but you would want to index the Var by the set of sets (which Pyomo currently does not support for a number of reasons).
That said, you can get (almost exactly) the behavior you want by indexing a Var with a non-finite Set:
>>> m = ConcreteModel()
>>> m.x = Var(PositiveIntegers, PositiveIntegers, dense=False)
>>> m = ConcreteModel()
>>> m.x = Var(NonNegativeIntegers, NonNegativeIntegers, dense=False)
>>> for i in range(2):
... for j in range(2):
... new_v = m.x[i,j] = 0
...
>>> m.x.pprint()
x : Size=4, Index=x_index
Key : Lower : Value : Upper : Fixed : Stale : Domain
(0, 0) : None : 0 : None : False : False : Reals
(0, 1) : None : 0 : None : False : False : Reals
(1, 0) : None : 0 : None : False : False : Reals
(1, 1) : None : 0 : None : False : False : Reals